REVISITING TECHNOLOGY ACCEPTANCE MODEL WITH SOCIAL INFLUENCE FACTORS

Transcription

1 REVISITING TECHNOLOGY ACCEPTANCE MODEL WITH SOCIAL INFLUENCE FACTORS Hee-Dong Yang, In-Young Choi Ewha Womans University Abstract Despite many numbers of studies in Technology Acceptance Model (TAM), the social influence construct has not been recognized, nor handled clearly. This study includes the social influence construct in TAM and tests the robustness of the model for different information systems contexts: office automation (spreadsheet) and Internet usage. Study findings include that 1) social influence has a direct impact on the usage of spreadsheet, but not on Internet usage, and 2) social influence has a stronger indirect impact on IS use through PEU than through PU (for both Internet and spreadsheet). Keyword: Technology acceptance model, Social Influence, Structural Equation Modeling 1. Introduction Studies and models are abundant in explaining IS acceptance or use. Due to the long history of this stream of research, several meta-models have been developed as the cumulative of previous studies: Lucas, Ginzberg & Schultz (1990), Davis (1989, 1993), and Moore & Benbasat (1991). Lucas, Ginzberg and Schultz (1990) developed and tested empirically the structural model of information systems implementation, distinguishing the models for managers and general users. In common, three factors were found important for the acceptance of information systems: decision style, job characteristics, and demographics. Organizational support was found important only for managers, whereas three other factors were important only for users: knowledge of system, assessment of system & support, and system characteristics. Technology Acceptance Model (TAM), based on the Theory of Reasoned Action (TRA: Ajzen and Fishbein, 1980), has been widely used for predicting the adoption and use of information technologies (Davis, Bagozzi, and Warshaw, 1989). The beliefs determining attitude and intention posited by TAM are perceived ease of use (PEU) and perceived usefulness (PU). Davis (1989, 1993) regressed onto these two important cognitive factors regarding IS acceptance. Numerous IS literature has followed to retest the construct validity of these two concepts, and confirmed it (e.g., Adams, Nelson & Todd, 1992; Chin & Todd, 1995; Hendrickson, Massey & Cronan, 1993; Subramanian, 1994). In his Technology Acceptance Model, Davis found that attitude was just a mediating factor of PU on IS use, and that social norm does not influence on IS use. Moore and Benbasat (1991) developed an instrument to measure the perceptual factors influencing information technology innovation, and identified eight important factors: voluntariness, relative advantage, compatibility, image, ease of use, result demonstrability, visibility, and trialability. These three meta-studies have different perspectives: Lucas, Ginzberg and Schultz (1990) take the organizational change perspective, Davis (1989, 1993) focuses on individual s perception of systems characteristics, and Moore and Benbasat (1991) take the innovation standpoint which concerns about the diffusion (spread) of technology across organization. 509

2 Commonly in these studies, social influence has not been recognized, nor handled clearly. Recently, the social influence on the adoption of information systems is getting more attention of researchers (e.g., Fulk, 1993; Markus, 1990; Orlikowski, 1992; Weick, 1990). However, it still remains as a subtle and vague construct. Therefore, we conduct an empirical study to make a clear understanding of this concept: i.e., what kind of social perceptual factors influence IS adoption, and how? Especially, this study focuses on tapping on TAM with social influence issues. This study contributes to the TAM research streams in the following couple of issues. First, this study includes social influence factors that have been disregarded in this research stream. TAM is an individual IS usage model. Identifying social influence on individual IS adoption or use could help improve the power of this model. Second, this study tests this expanded TAM model for two different information systems contexts: office automation (spreadsheet) and Internet usage (visiting and using web sites). Office automation software such as spreadsheet and word processor has been popular domains of TAM model research (e.g., Adams, Nelson & Todd, 1992; Chau, 1996; Davis et al., 1989; Hendrickson et al., 1993; Mathieson, 1991; Venkatesh & Davis, 1996). These products have alternative and competitive products, which means users may have chosen certain products voluntarily. However, it s not sure whether TAM can have robustness in explaining acceptance of Internet technology, which must be different from office automation tools. Internet is being used not only as an instrument to conduct tasks, but also as an instrument to communicate with virtual community relating to affective aspect of life (Armstrong & Hagel, 1996; Riggins & Rhee, 1999). Due to its exponential popularity and telecommunication characteristics, social or peer pressure may come to work on accepting Internet technology. Our study taps on the effectiveness of the expanded TAM model for both contexts of office automation and Internet. 2. Social Influence Factors We looked into the major theories regarding IS adoption (or use), such as TAM, TRA, TPB (Theory of Planned Behavior: Ajzen, 1985), the decomposed TPB (Taylor & Todd, 1995), and innovation theory (Moore & Benbasat, 1991), to identify what kind of constructs in these models can be grouped as social influence. Three constructs were found associated with social influence: subjective norm, visibility, and image. Among these three, the subjective norm of TRA is the representative construct in regards to social influence. Even though Davis et al. (1989) included the subjective norm in the original model as an external variable, they dropped it from the model due to the uncertain theoretical and psychometric status. Subjective norm is defined as the individual s perception of a referent other s opinion about the individual s performance of the behavior. (Fishbein and Ajzen, 1975, p.302). Subjective norm is determined by a multiplication of individual s normative beliefs and motivation to comply. Two other social influence factors can be identified from Moore & Benbasat s (1991) study: image and visibility. Image is defined as the degree to which adoption/usage of the innovation is perceived to enhance one s image or status in one s social system. (Moore & Benbasat, 1991: p.195). This is close to Chau s (1996) long-term PU and Tornatzky & Klein s (1982) social approval. Chau (1996) distinguished between near-term PU and longterm PU, stating the former relates to functionality of computer systems on improving job performance or enhancing job satisfaction, whereas the latter relates to improving one s social status. Tornatzky & Klein (1982) identified social approval as one of the ten characteristics addressed most frequently in over 100 innovation studies. Visibility means the degree to which the innovation is visible in the organization," which means the more a potential adopter can see an innovation, the more likely he is to adopt it 510

3 (Moore & Benbasat, 1991: p.195). Visibility is the very close concept to critical mass (Markus, 1990) and network externality (Riggins, Kriebel & Mukhopadhyay, 1994) that assert the usefulness of a network is primarily a function of the number of participants in the network. Compared to other social influence constructs, visibility is related to non-significant others because of the nature of critical mass theory. Therefore, we assume that these three constructs are different manifestations of social influence. 3. Research Model Figure 1 depicts the research model of this study. A couple of ponderous decisions were made a priori in this model. First, like other studies of TAM (e.g., Adams et al., 1992; Chau, 1996), the attitude construct is taken out to simplify the model. Second, IS use, not early adoption, is chosen for the dependent variable. The reason why we are more interested in IS use is that it may be the more appropriate measure of technology innovation diffusion rather than the early adoption or acquisition (Fichman & Kemerer, 1999). Also, our interest in the prediction of actual behavior adds to the reason why intention is dropped from the model. A number of TAM studies excluded attitude and used actual usage behavior instead of intention in their TAM studies (e.g., Adams et al. 1992, Gefen and Straub, 1997; Igbaria et al., 1995; Straub et al. 1995; Cheung et al. 2000). The following discussions will proceed with respect to each path of figure 1. [Figure 1] Research Model Social Norm Perceived of Ease of Use H5 Image Social influence H1 H6 H3 IS Use Visibility H2 Perceived of Usefulness H4 3.1 Impact of social influence on PU & PEU In TRA, all the external stimuli are assumed to influence behavior only indirectly via attitude or subjective norm (Fishbein & Ajzen, 1975: 396). TAM inserts two cognitive factors (PU and PEU) on the path between external stimuli and attitude, and insists that external stimuli influence a person's attitude toward a behavior indirectly by influencing PU and PEU. In TAM, external stimuli have been conceptualized diversely such as the various individual differences, situational constraints, and managerially controllable interventions (Davis et al., 1989: 988), as system design features (Davis, 1993: 476), or as the task, user characteristics, political influences, organizational factors, and the development process (Szajna, 1996: 86). TAM is focused on individual psychological status. Moving beyond a concern with one user and an interface, the socio-technical perspective argues that a network of social relationships surrounds all working practices (e.g., Orlikowski, 1992; Weick, 1990). Technology historian Carl Mitcham (1994) recognizes the "unintended intentions" (symbolic import) in individual's adoption of technology artifacts, which address various styles and usage. This perspective 511

4 insinuates that individual s beliefs are socially constructed through interactions with other members. Therefore, the social influence must have significant impact on the individual PU and PEU over the characteristics of computer systems. H1: Social influence will have significant impact on determining PEU of a system. H2: Social influence will have significant impact on determining PU of a system. 3.2 Relationship between PEU and PU Davis (1993) provides a theoretical rationale regarding the relationship between PU and PEU as follows: PU concerns the expected overall impact of system use on job performance (process and outcome), whereas PEU pertains only to those performance impacts related to the process of using the system per se. Consider a hypothetical new forecasting model, which although equally easy to use as the model it supersedes, provides a more accurate forecast. Moving from the old model to the new one, PU goes up with no effect on PEU. Thus, PEU influences PU but not vice versa. (Davis, 1993: 477-8). This theoretical relationship has been supported by empirical evidences (e.g., Chau, 1996; Davis, 1989, 1993; Davis et al., 1989; Mathieson, 1991; Szajna, 1996; Taylor & Todd, 1995; Venkatesh & Davis, 1996). Consistent with these theoretical and empirical arguments, our study also assumes that PEU influences PU, not vise versa. H3: Perceive ease of use of a system will have significant impact on determining PU of a system. 3.3 Influence of PU & PEU on IS Use Davis (1989) assumed that PEU does not influence IS use, even though it influences PU, stating PEU may actually be a causal antecedent to PU, as opposed to a parallel, direct determinant of system usage. (p. 319). Adams, Nelson & Todd (1992) warned not to focus on PEU in identifying the antecedents to IS use, saying A heavy emphasis on PEU, particularly at the cost of functionality, is not advisable (p. 237). Keil, Beranek and Konsynski (1995) nixed PEU-focused efforts such as developing a good user interface, stating No amount of PEU will compensate for low usefulness (p. 89). Even though the influence of these two constructs (PU & PEU) on IS use has not been consistent in empirical studies, most of TAM research share the common conclusion: PU influences on use, whereas PEU does not (Chau, 1996; Davis et al. 1989; Davis, 1989, 1993; Straub et al. 1997; Szajna, 1996). Our study takes this majority s opinion that PU leads to use, whereas PEU does not. H4: PU of a system will have significant impact on system use. H5: PEU of a system won t have significant impact on system use. 3.4 Impact of Social Influence on IS Use Adams, Nelson & Todd (1992) and Subramanian (1994) reported that PU and PEU explain only marginal variances of IS use. In their studies, those two factors explain around 30% of the variance of IS use. These results imply that there must be additional factors that influence individual usage of IS. We can find a possible answer (i.e., missing factor) from other IS use 512

5 theories: TRA, TPB, and Innovation perspective. In common, these alternative theories include the "social influence" factors as antecedents of individual IS use. TRA and TPB contain "subjective norms" in their models. Moore & Benbasat's (1991) Innovation perspective includes two other social influence constructs, visibility and image. Meanwhile, the debate regarding the direct influence of social factors on IS use has not been clearly settled. TRA, TPB, and Decomposed TPB in common assumed and empirically found that subjective norms are directly linked to IS use or adoption. Numerous studies also reported subjective norm is an important determinant of IS adoption or use (e.g., Cooper & Zmud, 1990; Hartwick & Barki, 1994; Laudon, 1985). Some studies delicately tested the influence of subjective norm by differentiating IS adoption and continuous use in terms of chronological diffusion of technology, and identified subjective norm influences only on early IS adoption (Hartwick & Barki, 1994; Karahanna, Straub & Chervany, 1999). On the contrary, Davis, Bagozzi and Warshaw (1989) and Mathieson (1991) reported non-significant path from subjective norms to the intention of IS use. Our study includes visibility and image with subjective norm as social influence constructs and tests if the expanded TAM has significant impact on system use. H6: Social influence will have significant impact on system use. Hypothesis 3, 4, and 5 may not sound new because previous TAM studies empirically conducted numerous hypotheses regarding PU and PEU, including the relationship between PU and PEU, and the relative importance of PU and PEU on IS use. Structural equation modeling estimates a series of separate, but interdependent, multiple regression equations simultaneously (Hair, Anderson, Tatham and Black, 1998). Therefore, the inclusion of social influence factors onto TAM must influence the strength of paths surrounding PU and PEU. 4. Methodology 4.1 Data Collection Data was collected from under-graduate students who major in MIS (management information systems) in a college of management in New England area. Samples are regarded homogeneous in terms of demographic features such as age, grade and major. Students who were asked for survey already took or were taking the MIS courses such as the introduction to computers, database management, business network, programming languages. The questionnaires were handed out purposely late November, so that students are familiar with the MIS courses that they were taking. Students were asked to fill out both spreadsheet and Internet usage surveys anonymously and submit them to the class instructors on the voluntary basis. It took 9 weeks to finish collecting surveys. In total, 211 valid questionnaires were returned for spreadsheet usage, and 206 perfect questionnaires were returned for Internet usage out of 420 handouts, recording 50.2% and 49.0% of return ratio, respectively. 4.2 Questionnaire All of measurement items are taken from the related research. Davis, Bagozzi & Warshaw (1989) measurement items regarding PU, PEU, and system use are taken. Measurement items of subjective norms came from Mathieson (1991) and Taylor & Todd (1995). Measurement items of visibility and image came from Moore and Benbasat (1991). Among the original 513

6 four items for the visibility scale, the statement PWS are not very visible in my organization is excluded because the word visible is used redundantly to measure visibility. 4.3 Assessment of reliability and validity The reliabilities for each scale are shown in Table 1. The measurement scales used in the study show high levels of reliability - Cronbach s alpha is above 0.60 for both software use and Internet use. [Table 1] Reliability Estimates Construct Items Spreadsheet Internet Cronbach s alpha Cronbach s alpha PU PEU Image Visibility Subjective Norm IS Use Convergent validity is assessed by factor analysis of the scales. The factor loadings are shown in Table 2. Through Varimax rotation, the 25 items are cleanly loaded onto 6 factors - PEU, PU, image, visibility, subject norm and IS use. [Table 2] Factor Analysis of Scales Spreadsheet Internet Fac.1 Fac.2 Fac.3 Fac.4 Fac.5 Fac.6 Fac.1 Fac.2 Fac.3 Fac.4 Fac.5 Fac.6 PU PU PU PU PEU PEU PEU PEU IMG IMG IMG IMG VS VS VS SN SN

7 USE USE Notes: PU = Perceived Usefulness, PEU = Perceived Ease of Use, IMG = Image, VS = Visibility SN = Social Norm We assess the discriminant validity by testing if all items load more highly on their associated construct than on any other constructs. The square root of the average variance extracted for each construct was compared to the correlations between each construct and other constructs. All the values of the square root of the average variance (i.e., the diagonals) were greater than the correlations between constructs (i.e., the off-diagonals) in table 3, indicating that all the constructs in the model exhibited the discriminant validity. [Table 3] Discriminant Validity Spreadsheet Internet PU PEU IMG VS SN USE PU PEU IMG VS SN USE PU PEU IMG VS SN USE Note: Diagonal elements are the square roots of the average variance extracted and off-diagonal elements are correlations between constructs. 5. Data Analysis Our empirical test would consist of two phases. The first phase is to check the second-order latent factor structure. This test is about whether those three first order latent factors are regressed onto the higher level latent factor. If the first-order latent factors regress onto the second-order factor, this higher-level factor could be named as the general "social influence." The second phase is to test the research model of figure 1. These steps of statistical analysis will be applied for both spreadsheet usage and Internet usage of students. 5.1 Test of Social Influence Factor Structure By the review of various IS implementation literature, we proposed that social influence context might have a second-order factor structure with three dimensions - subjective norm, image and visibility. To test the proposed multi-dimensionality of the social influence construct, we conduct the confirmatory factor analysis using AMOS. The value of χ 2 for overall model fit with 24 degrees of freedom is significant for both cases. However, The chisquare test has been recognized as an inappropriate test for large sample sizes (Browne & Cudeck, 1993, Marsh, 1994). As recommended by Hair et al. (1998), several other fit indices are examined. The Goodness of Fit Index (GFI) and the adjusted Goodness of Fit Index (AGFI) should be greater than 0.90 and 0.80, respectively, and the Root Mean Square Error of Approximation (RMSEA) should be lower than 0.08 (Joreskog and Sorbom, 1993; Browne & Cudeck, 1993). The values of GFI, AGFI and RMSEA indicate a marginally satisfactory fitness for both spreadsheet and Internet data [Table 5]. To improve the model fitness, we refer to the modification indices that address the improvement of model fitness by adding 515

8 certain non-estimated paths. The value of modification index corresponds approximately to the reduction of chi-square if the path is included in the model. The criterion value to include the path into the model is 3.84 or greater (Hair et al., 1998). This kind of modification is sometimes recommended to improve model fitness in structural equation model (Hair et al., 1998; Arbuckle, 1997; Bone, Sharma & Shimp, 1989). The current modification indices from our dataset suggest allowing the correlation between er1 and er2. This modification is acceptable because the correlation between er1 and er2 is the highest among all the image measurement items [Table 4]. This change helped all the goodness-of-fit indices surpass the acceptable level as shown in Table 5. Thus, the proposed hierarchical structure is very strongly supported [Figure 2]. [Table 4] Correlation Matrix of Image items Spreadsheet Internet IMG1 IMG2 IMG3 IMG4 IMG1 IMG2 IMG3 IMG4 IMG IMG2.766 ** ** 1.00 IMG3.444 **.535 ** **.642** 1.00 IMG4.627 **.670 **.638 ** **.760**.825** 1.00 ** p<0.01 [Table 5] Summary Results for Confirmatory Factor Analysis for the Proposed Factor Structure of Social Influence Fit Index Spreadsheet Internet Initial model Revised model Initial model Revised model χ 2 (DF) (24) (23) (24) (23) P value(>0.05) χ 2 /df(<5) GFI(>0.90) AGFI(>0.80) RMSEA(<0.08) [Figure 2] Social Influence as a Second-Order Factor Structure Social Influence er10.67 er er12 Image Critical Mass Social Norm IMG1 IMG2 IMG3 IMG4 VS1 VS2 VS3 SN1 SN2 er1 er2 er3 er4 er5 er6 er7 er8 er9.50 a) Spreadsheet 516

9 Social Influence er10.39 er er12 Image Critical Mass Social Norm IMG1 IMG2 IMG3 IMG4 VS1 VS2 VS3 SN1 SN2 er1 er2 er3 er4 er5 er6 er7 er8 er9.59 b) Internet 5.2 Test of the Proposed TAM model Structural equation analysis was conducted using Amos to test the research model of figure 1. There is no single recommended measure of model fitness. Therefore, a variety of measures is suggested (Bentler and Bonett, 1980; Fornell, 1983; Hayduk, 1987). A chi-square analysis indicates that the model for spreadsheet does not fit the data (Chi-square=65.512:p<0.005), but the model for Internet fits the data (Chi-square=39.175:p>0.005). The chi-square test has been recognized as an inappropriate test for large sample sizes (Browne & Cudeck, 1993, Marsh, 1994). Instead, Chi-square divided by degrees of freedom is suggested, and the value below 5 is recommended as the criterion (Wheaton, Muthen, Alwin and Summers, 1977). The value of Chi-square divided by degrees of freedom is approximately 1.7 for spreadsheet and 1.0 for Internet, which is below the cutoff-value. As mentioned above, the values of GFI, AGFI and RMSEA are examined for the judgment of the fitness of the proposed model. In our dataset, GFI is and and AGFI is and 0.943, respectively for spreadsheet and Internet. RMSEA is at the acceptable level with and 0.012, respectively for spreadsheet and Internet. Together, these four measures indicate that our model fits to data for both spreadsheet and Internet [Table 6]. [Table 6] Fit Measures for the proposed TAM model Spreadsheet Internet DF χ P value(>0.05) p=0.004 p=0.417 χ 2 /df(<5) GFI(>0.90) AGFI(>0.80) RMSEA(<0.08)

10 5.3 Results and Analysis Analysis results of our hypotheses are summarized in figure 3 and 4, and table 7. The results are not entirely consistent across technologies, nor are they consistent with previous findings. [Table 7] Summary of Results (path coefficients) Spreadsheet Remarks Internet Remarks H1 SI -> PEU 0.56(0.000)** Supported 0.52(0.000)** Supported H2 SI -> PU 0.49(0.000)** Supported 0.51(0.000)** Supported H3 PEU -> PU 0.18(0.047)* Supported 0.14(0.123) Rejected H4 PU -> USE 0.09(0.390) Rejected 0.03(0.812) Rejected H5 PEU -> USE 0.23(0.02)* Rejected 0.419(0.001)** Rejected H6 SI -> USE 0.27(0.046)* Supported 0.07(0.600) Rejected * <0.05 ** <0.01 [Figure 3] Path Coefficients of Spreadsheet Usage PEU1 PEU2 PEU3 PEU SN IMG Social Influence.56** PEU.18*.27*.23* USE.82 USE1 VS.57.49** PU USE PU1 PU2 * <0.05 ** <0.01 [Figure 4] Path Coefficients of Internet Usage PEU1 PEU2 PEU3 PEU SN IMG Social Influence.51** PEU ** USE.59 USE1 VS.47.51** PU USE PU1 PU2 * <0.05 ** <

11 The path from social influence to PU and PEU is significant for both spreadsheet (0.56(0.000) and 0.49(0.000)) and Internet (0.52(0.000) and 0.51(0.000)). Therefore, hypothesis 1 and 2 are supported for both spreadsheet and Internet: i.e., the social influence has substantial impact on PEU and PU. Figure 3 and 4 show that the path from PEU to PU is significant only for spreadsheet (0.18(0.047)). This result reconfirms the results of the previous TAM research (Hartwick & Barki, 1994; Taylor & Todd, 1995). However, the path is not significant for Internet usage. Therefore, hypothesis 3 is partially supported. We get the reverse results about the hypothesis 4 and 5 for both spreadsheet and Internet: i.e., PEU influences IS use, whereas PU doesn t. The above results regarding hypothesis 3, 4, and 5 indicate that PEU has direct influence on spreadsheet usage as well as indirect one through PU, whereas it has only direct influence on Internet usage. These inconsistent findings may pertain to the nature of the technology itself that will be discussed in the next section. The relative importance of PU and PEU on IS use has been disputable in numerous MIS studies. The detailed discussion on this issue will also follow in the next section. The path from social influence to IS use is significantly supported only for spreadsheet usage (0.027(0.046)). Therefore, hypothesis 6 is partially supported. These inconsistent findings may pertain to the nature of the technology itself that will be discussed in the following section. 6. Discussion This section discusses a couple of major unexpected results of our study: i.e., the relative importance of PU and PEU on IS use, and the different nature of spreadsheet and Internet. The first issue pertains to the test results of hypothesis 4 and 5, and the second one is associated with those of hypothesis 3 and 6. Even though our results regarding the relative importance of PU and PEU on IS use (hypothesis 4 & 5) are at odds with the original TAM study, we can notice several TAM studies reported the similar results as ours. For example, Adams, Nelson & Todd (1992) found that the path from PEU to IS use was significant from their cross-sectional study that surveyed experienced users for their spreadsheet software usage. They reported that the experienced users place less weight on PU rather than PEU. The context of their study was similar to ours. Igbaria, Zinatelli, Cragg & Cavaye(1997) reported that, in small firms, PEU influence IS usage more significantly than PU. Thompson, Higgins & Howell(1991) also found the near-term complexity, which is similar to PEU, has significant path coefficient to PC utilization. Lu & Gustafson (1994) reported that PU influence neither early adoption nor stable use, whereas PEU influences both. In addition to the empirical history above, the path from PEU to IS use has theoretical rationale. Bandura (1997) identifies two separate expectations affecting individual behaviors: self-efficacy and outcome expectancy. Self-efficacy refers to an individual s belief that he/she possesses the requisite skills and abilities to accomplish an identifiable task, whereas outcome expectancy refers to a belief of outcome values, so that individuals are likely to undertake behaviors they believe will result in valued outcomes. Bandura states that selfefficacy typically has a larger effect, and has also a direct impact on outcome expectancy. In the IS context, self-efficacy is closer to PEU, whereas outcome-expectancy to PU (Compeau & Higgins, 1995; Davis, 1989; Lu & Gustafson, 1994). Some regarded self-efficacy as an antecedent to PEU (Igbaria, Zinatelli, Cragg & Cavaye, 1997; Venkatesh & Davis, 1996), and related to various aspects of computer usage (Compeau & Higgins, 1995). All these empirical findings and theoretical arguments imply that PEU could influence IS use more than PU does. 519

12 The split support for hypothesis 3 and 6 can be attributable to the nature of the technology itself. While the spreadsheet is task-oriented, the Internet is for other personal purposes such as fun and sympathy. Although the use of spreadsheet is not compulsory for these users, it has become a de facto standard for performing task. Tayor and Todd (1995) reported that in deciding whether to use the facility, students are likely to be influenced by both what their professors might think, due to possible impact on their grades, and by what their peers think due to the competitive nature of the environment. Though the spreadsheet usage is not compulsory, the social influence on the usage of spreadsheet has been recognized elsewhere (e.g., Moore and Benbasat, 1991; Adams et al., 1995). However, Internet is being used to communicate with virtual community or search useful information (Armstrong & Hagel, 1996; Riggins & Rhee, 1999). Thus, the spreadsheet is more attentive to social cues while Internet attends to rather affect-related stimuli such as PEU. 7. Conclusions The purposes of our study were (a) to empirically examine the indirect and direct impact of social influence in technology acceptance model and (b) to examine if the expanded TAM is applicable for two different information systems contexts: spreadsheet and Internet. The findings to these questions are as follows. First, we found that social influence, which has been disregarded in TAM, is an important determinant in explaining technology acceptance and use of spreadsheet. In both cases of spreadsheet and Internet use, social influence is found to have substantial impact on both PEU and PU, and has a stronger indirect impact on IS use through PEU than PU. Second, we also found that the direct or indirect impact of social influence is different between technologies. The different results may pertain to the different nature of technologies. While the spreadsheet is task-oriented, the Internet is for other personal purposes such as fun and sympathy. So, the usage of Internet is relatively less compulsive than spreadsheet. The current study provides preliminary evidence suggesting that technology acceptance and usage behavior are determined by social influence factors as well as individual perception factors in a certain technology. Our study negates the definitive perspectives (both positivists and negativists) regarding the social influence on IS usage. We argue that social influence has differential impacts on IS use according to the characteristics of information technology. Given this is just a single study, our results should be replicated by future studies before we suggest definitive changes in the existing theories of technology acceptance. Like other social studies, the current study has several limitations. First, we used only perceptual measures of IS use. Many studies have shown that individuals perceptions of IS usage are sometimes quite different from their actual usage pattern (Collopy, 1996; DeLone & McLean, 1992; Straub, Limayem, Karahanna-Evaristo, 1995; Tricy & Treacy, 1986). Second, our sample was from undergraduate students who major in MIS. They may be more knowledgeable (therefore, spoiled) about computer features. Therefore, it is difficult to apply the result of this study to inexperienced end users. Also, no organizational setting is considered in our data set. IS implementation studies have emphasized the importance of organization issues (such as structure, size, industry, support, change management, etc.). These issues could not be considered in this study at all. Third, since we used a cross-sectional data set, with both independent and dependent variables collected contemporaneously, the results are susceptible to same method bias. That is, the correlation between independent and dependent variables, and the explained variance 520

13 of the dependent variable, might have been inflated because these were all measured at the same time within the same questionnaire. References Adams, D.A., Nelson, R.R. and Todd, P. A. PU, Ease of Use, and Usage of Information Technology: A Replication, MIS Quarterly, June, pp , 1992 Ajzen, I. and Fishbein, M. Understanding Attitudes and Predicting Social Behavior Prentice- Hall, Englewood, NJ, 1980 Ajzen, I. From Intentions to Actions: A Theory of Planned Behavior. In Kuhl, J. & J. Beckman (eds.), Action Control: From Cognition to Behavior, Springer Verlag, New York, 1985 Arbuckle, James L. Amos Users Guide Version 3.6, SmallWaters Co., Chicago, 1997 Armstrong, A. and Hagel, J. The Real Value of On-line Communities Harvard Business Review, Vol.74, No.3, pp , 1996 Bandura, A. Self-Efficacy:The Exercise of Control, New York: W.H. Freeman, Bentler, P. and Bonnett, D., Significance Tests and Goodness of Fit in the Analysis of Covariance Structure, Psychological Bulletin, Vol.88, No.3, pp , 1980 Browne, M. W., and Cudeck, R. Alternative ways of assessing model fit. In Bollenk, K. A., Long, J. S. (Eds.) Testing structural equation models, California: Sage, pp , 1993 Bone, P. F., Sharma, S., and Shimp, T. A. A Bootstrap Procedure for Evaluating Goodnessof-Fit Indices of Structural Equation and Confirmatory Factor Models, Journal of Marketing Research, Vol. 16, Feb., pp , 1989 Chau, P.Y.K. An Empirical Assessment of a Modified Technology Acceptance Model Journal of Management Information Systems, Vol.13, No.2, pp , 1996 Chin, W.W. and Todd, P.A. On the Use, Usefulness, and Ease of Use of Structural Equation Modeling in MIS Research: A Note of Caution, MIS Quarterly, Vol.19, No.2, pp , 1995 Collopy, F. Biases in Retrospective Self-Reports of Time Use: An Empirical Study of Computer Users, Management Science, Vol.42, No.5, pp , 1996 Compeau, D. R. & Higgins, C. A. Computer Self-Efficacy: Development of a Measure and Initial Test, MIS Quarterly, Vol.19, No.2, pp , 1995 Cooper, R.B. and Zmud, R.W. Information Technology Implementation Research: A Technological Diffusion Approach, Management Science, Vol.36, No.2, pp , 1990 Davis, F.D. PU, PEU, and User Acceptance of Information Technology, MIS Quarterly, Vol.13, No.3, pp , Davis, F.D., Bagozzi, R. and Warshaw, P.R. User Acceptance of Computer Technology: A Comparison of Two Theoretical Models Management Science, Vol.35, No.8, pp , 1989 Davis, F.D. User Acceptance of Information Technology: System Characteristics, User Perceptions and Behavioral Impacts, International Journal of Man-Machine Studies, Vol.38, No.3, pp ,

14 DeLone, W. H. and McLean, E. R. Information Systems Success: The Quest for the Dependent Variable, Information Systems Research, Vol.3, No.1, pp.60-95, 1992 Fichman, R.G. & C.F. Kemerer. The Illusory Diffusion of Innovation: An Examination of Assimilation Gaps, Information Systems Research, Vol.10, No.3, pp , 1999 Fishbein, M. & Ajzen, I. Beliefs, Attitude, Intention, and Behavior: An Introduction to Theory and Research. Addison-Wesley, Reading, MA Fornell, C., Issues in the Application of Covariance Structure Analysis: A Comment, Journa of Consumer Research, March, Vol.9, No.3, pp , 1983 Fulk, J. Social Construction of Communication Technology, Academy of Management Journal, Vol.36, No.5, pp , 1993 Hair, J. F., Anderson, R. E., Tatham, R. L. and Black, W. C. Multivariate Data Analysis (5th edition), Prentice Hall, Upper Saddle River, NJ, 1998 Hartwick, J. and Barki, H. Explaining the Role of User Participation in Information System Use, Management Science, Vol.40, No.4, pp , 1994 Hayduk, L. A. Structural Equation Modeling with LISREL, Johns Hopkins University Press, Baltimore, MD Hendrickson, A.R., Massey, P.D. and Cronan, T.P. On the Test-Retest Reliability of PU and PEU Scales, MIS Quarterly, June, pp , 1993 Igbaria, M., Zinatelli, N., Cragg, P. & Cavaye, A.L.M. Personal Computing Acceptance Factors in Small Firms: A Structural Equation Model, MIS Quarterly, Vol.21, No.3, pp , 1997 Joreskog, K. G., and D. Sorbom, LISREL 8: Structural Equation Modeling with the SIMPLIS Command Language, Mooresville, III.: Scientific Software, Karahanna, E., Straub, D. and Chervany, N.L. Information Technology Adoption across Time: A Cross-Sectional Comparison of Pre-Adoption and Post-Adoption Beliefs, MIS Quarterly, 23(2), pp , 1999 Keil, M., Beranek, P.M. and Konsynski, B.R. Usefulness and Ease of Use: Field Study Evidence Regarding Task Considerations, Decision Support Systems, Vol.13, No.1, pp.75-91, 1995 Laudon, K.C. Environmental and Institutional Models of Systems Development: A National Criminal History System, Communications of the ACM, 28(7), 1985, pp Lu, H.P. & Gustafson, D. H. An Empirical Study of PU and PEU on Computerized Support System Use over Time, International Journal of Information Management, 14, 6, 1994, pp Lucas, H. C. Jr, Ginzberg, M.J. and Schultz, R. L. Information Systems Implementation: Testing a Structural Model, Norwood, NJ: Ablex Publishing, Markus, M.L. Toward a Critical Mass Theory of Interactive Media. In J. Fulk & C. Steinfield (eds.), Organizations and Communication Technology, Newbury Park, CA: Sage, Marsh, Hert W., Confirmatory Factor Analysis Models of Factorial Invariance: A Multifaceted Approach, Structural Equation Modelling, Vol.1, No.1, pp.5-34, 1994 Mathieson, K. Predicting User Intentions: Comparing the Technology Acceptance Model with the Theory of Planned Behavior Information Systems Research, Vol.2, No.3, pp ,

15 Mitcham, C. Thinking Through Technology: The Path between Engineering and Philosophy, The University of Chicago Press, Chicago, IL, Moore, G.C. and Benbasat, I. Development of an Instrument to Measure the Perceptions of Adopting an Information Technology Behavior, Information Systems Research, Vol.2, No.3, pp , 1991 Orlikowski, W. J. The Duality of Technology: Rethinking the Concepts of Technology in Organizations, Organization Science, Vol.3, No.3, pp , 1992 Riggins, F.J. and Rhee, H. S. Developing the Learning Network Using Extranets, International Journal of Electronic Commerce, Vol.4, No.1, pp.65-83, 1999 Riggins, F.J., Kriebel, C.H. and Mukhopadhyay, T. The Growth of Interorganizational Systems in the Presence of Network Externalities, Management Science, Vol.49, No.8, 1994, pp , 1994 Straub, D., Keil. M. and Brenner, W. Testing the Technology Acceptance Model across Cultures: A Three Country Study, Information & Management, 33, pp.1-11, 1997 Straub, D., M. Limayem & E. Karahanna-Evaristo Measuring System Usage: Implications for IS Theory Testing, Management Science, 41(8), , 1995 Subramanian, G. H. A Replication of PU and PEU Measurement, Decision Sciences, 25(5/6), pp , 1994 Szajna, B. Empirical Evaluation of the Revised Technology Acceptance Model, Management Science, Vol.42, No.1, pp.85-92, 1996 Taylor, S. and Todd P.A. Understanding Information Technology Usage: A Test of Competing Models, Information Systems Research, Vol. 6, No.2, pp , 1995 Thompson, R. L., Higgins, C. A. and Howell, J. M. Personal Computing:Toward a Conceptual Model of Utilization, MIS Quarterly, March, pp , 1991 Tornatzky, L. and Klein, K. Innovation Characteristics and Innovation Adoption Implementation: A Meta-Analysis of Findings, IEEE Transactions on Engineering Management, 29(1), pp.28-45, 1982 Trice, A.W. and Treacy, M.E Utilization as a Dependent Variable in MIS Research, Proceedings of the International Conference of Information Systems, pp , 1986 Venkatesh, V. and Davis, F. D. A Model of the Antecedents of PEU: Development and Test, Decision Sciences, 27(3), pp , 1996 Venkatesh, V. and Morris, M. G. Why Don t Men Ever Stop to Ask for Directions? Gender, Social Influence, and Their Role in Technology Acceptance and Usage Behavior, MIS Quarterly, 24, 1, pp , 2000 Weick, K. E. Technology as Equivoque: Sensemaking in New Technologies, In P.S. Goodman, L.E. Sproull & Associates (eds.), Technology and Organizations, San Francisco: Jossey-Bass, pp.1-44, 1990 Wheaton, B., D. Muthen, D. Alwin, & Summers, G. Assessing Reliability and Stability in Panel Models, in D. Heise (ed.), Sociological Methodology, San Franciscon, CA: Jossey- Bass,